Men Explain "Men Explain Things to Me"

Three-Channel Audio:

CLICK HERE for each of the audio snippet.

Motivation

NotebookLM, developed by Google, is an AI-powered product designed to generate audio overviews, offering podcast-like conversations based on uploaded materials. By default, the host voices include one man and one woman, providing a balanced conversational dynamic. While the product itself is impressive, I found myself uncomfortable at times, particularly when the male voice dominated the conversation with explanations, leaving the female voice relegated to filler words or reactive comments. This artistic project stems from that discomfort and aims to delve into two core themes:

1. The Signal-to-Noise Dilemma in AI-Generated Media: As generative AI becomes increasingly prolific, we face an escalating problem of content saturation, reducing the signal-to-noise ratio on the internet. The mixed audio in this project is intentionally overwhelming to replicate the sensation of being inundated with information.
2. Gendered Perspectives in AI-Generated Conversations: The audio samples are derived from NotebookLM’s overviews of three seminal feminist texts:
   • Men Explain Things to Me by Rebecca Solnit
   • The Second Sex by Simone de Beauvoir
   • A Room of One’s Own by Virginia Woolf

By analyzing these audio overviews, I sought to interrogate how generative AI handles gendered experiences. Specifically, I isolated instances of “mansplaining” or when the male voice addressed female experiences in ways that felt incongruous or reductive.

The Process

Listening to NotebookLM’s audio outputs, I cut and curated sections that highlighted imbalances in dialogue dynamics. For example, in “audio clip 10,” the male voice describes the experience of being a young woman transitioning from childhood to adulthood, stating: “I think every woman remembers it. You’re caught between childhood and adulthood trying to figure out who you are, where you fit in, while dealing with all this societal pressure.”—a curious and somewhat jarring choice considering the subject matter. However, in testing, I discovered that NotebookLM’s assignment of voices is largely randomized, and the “meatier” content doesn’t always come from the male voice. This variability raises thought-provoking questions:

• In discussions of feminist or gender-specific topics, is there ever a need for a male voice?
• Conversely, in traditionally male-dominated subject areas, how would an all-female voice cast be received?
• Should future generative AI tools allow users to customize the gender, race, or other identity markers of the “speakers”?

The Artistic Statement

The final audio pieces are a deliberate cacophony, mirroring the cognitive overload we experience in a world saturated with generative content. By pairing these overwhelming soundscapes with feminist writings, the work invites listeners to reflect on the intersections of voice, authority, and representation in AI media. The discomfort of “mansplaining”—a term coined to describe patronizing explanations by men to women—is amplified here as an artistic critique.

Disclaimer

This project involves heavy editing for artistic purposes and does not represent NotebookLM’s audio generation capabilities as a whole. Instead, it highlights specific moments to provoke thought and discussion.

Future Directions: The AI Bechdel Test

Inspired by the Bechdel Test in film, which examines the representation of women, a potential next step is developing an “AI Bechdel Test.” This could assess whether AI-generated dialogues meet certain standards for inclusivity and meaningful representation. For instance:

• Does the conversation include voices of diverse genders?
• Are those voices given equal weight in delivering substantive content?
• Are identity markers customizable to reflect varied user preferences and contexts?

Conclusion

This project is a starting point for exploring the deeper implications of AI’s role in shaping our media landscapes. As we design future generative tools, let’s consider not only the content but the voices through which that content is delivered. Representation, balance, and intentionality matter, even in AI-generated conversations.